This invention relates to a method and apparatus for depositing material on a substrate using a chemical vapor deposition process. More particularly, the invention relates to such a method and apparatus for producing multilayered structures of extreme thinness and sharp transitions between layers.
Chemical vapor deposition processes have long been used for depositing material on a substrate, such as in the fabrication of semiconductor devices. One such process, metalorganic chemical vapor deposition (MOCVD), is preferred for the fabrication of many high performance electronic and optoelectronic devices because it can utilize aluminum, which the other processes cannot do. In addition, the inherent characteristics of metalorganic chemical vapor deposition make it ideally suited for growing epitaxial layers, which are inevitably involved in new device structures.
In the growth process with metalorganic chemical vapor deposition, one or more film constituents are transported to a reaction zone in a form of gaseous reactants. The metalorganic reactant is contained in a bubbler through which a carrier gas is bubbled to vaporize the metalorganic for transport to the reaction zone. The desired compound forms as a layer on the substrate via the pyrolysis of the reactants and the subsequent recombination of the atomic or molecular species on the heated substrate. The growth process can be controlled by fixing the flow rates and thereby relative proportions of the various gaseous reactants with electronic mass flow controllers. Complex, multilayer epitaxial structures are formed by exchanging one gas composition for another using conventional gas-mixing systems.
Current chemical vapor deposition apparatus, commonly known as reactors, include a deposition chamber or vessel for housing the substrate, a gas mixing manifold, and a heat source such as an inductively coupled RF generator. The chamber is typically one of two configurations: a vertical bell jar usually employed for commercial, high volume production and a horizontal cylindrical tube preferred for research. A graphite susceptor for supporting the substrate is mounted in the chamber for positioning the substrate at a desired angle to the reactive gas flow. Examples of such chambers are disclosed in U.S. Pat. Nos. 3,306,768 to Peterson, 3,850,679 to Sopko et al., 4,066,481 to Manasevit, 4,369,031 to Goldman et al., 4,446,817 to Crawley, and in P. Daniel Dapkus, "Metal Organic Chemical Vapor Deposition," Annual Review of Material Sciences 1982, Vol. 12, pp. 243, 269.
One drawback of these conventional chamber designs is their inability to concentrate the flow of reactants near the surface of the substrate. The relatively large volume of the chamber encourages a turbulent flow of the gases from the chamber's inlet to its outlet, leading to a nonuniform deposition. The large volume also results in a low velocity of gas flow across the substrate, causing further nonuniformity in the deposition as the reactants deposit at the leading edge of the substrate.
The second component of most deposition apparatus, the gas-mixing manifold, directs gas flow from a number of sources into the chamber. The conventional manifold is typically a linear array of gas flow lines constructed from stainless steel tubing, connecting the various gas sources to the chamber inlet. Inlet valves are positioned in the flow lines for sequentially introducing different reactive gases into the chamber. An electronic mass flow controller controls the flow of gas in each line. The mass flow rates and sequencing of the valves are controlled by a central processor, which automates the growth of complex multilayer material structures.
The conventional manifold, although adequate for producing less advanced devices, cannot control reactive gas flows accurately enough to produce the sharp transitions in composition between layers demanded by today's high speed solid-state devices. The reactive gases tend to linger in any "dead space" or volume of the manifold downstream of the respective inlet valve. These gases gradually disperse into the chamber to contaminate the following deposition layer. The result is a gradual rather than sharp transition between layers. The manifold of Peterson, for example, employs flow meters and valves to control each reactive gas flow, but between the valves and the process chamber is a large volume from which reactants can disperse into the chamber even after the flow is shut off by the flow meter. The same drawback is apparent in other apparatus such as that of Sopko et al. and Manasevit, both of which use a linear manifold for injecting the reactive gases into the chamber. The inlet valve in each gas line is located a different distance from the outlet of the manifold, compounding the problem because the volume of "dead space" for each reactive gas differs. Other manifolds with similar drawbacks are disclosed in Goldman et al., Crawley, U.S. Pat. Nos. 4,279,670 to Steele, and 4,476,158 to Baumberger et al.
Much development in chemical vapor deposition has concentrated on improving the deposition method. Peterson, one of the earliest patents in the field, is directed to a process for coating a substrate with an oxide film while keeping the substrate at a relatively low temperature. Sopko et al. is directed to a method of coating a substrate by vaporizing the reactant and directing it through a nozzle against the substrate.
These and the other methods, however, have not been able to control precisely the uniform thickness of the deposited layers because they do not maintain a constant flow and thus a constant pressure through the process chamber. Constant flow and pressure are difficult to maintain in the face of the continuous switching of different reactants into and out of a chamber. Each reactive gas may have a different flow rate requiring compensating change in flow elsewhere to maintain a constant total gas flow into the process chamber. Pressure can also build in gases that are held behind closed valves, causing flow surges when the gas is eventually switched into the chamber. This pressure buildup is especially a problem in the MOCVD process, where pressure buildup within the bubbler can cause large bubbles to form and disrupt the reactive gas flow. The apparatus disclosed in Manasevit minimizes this pressure surge somewhat by directing gas flows to an exhaust rather than terminating them when they are not called for in the deposition process. However, Manasevit does not compensate for the effect of a varying reactive gas flow on the uniformity of deposition.